Project Summary/Abstract Frail elders living at home are a population at risk for decline and death as they often have multiple chronic conditions and complex social needs that may lead to unnecessary hospitalizations or potentially futile treatment at the end of life thereby driving up individual and public socioeconomic burdens. Advanced care planning, a process in which preferences for care at the end of life are discussed and placed in writing, is one approach to ensure that patient-centered decision making drives end of life care. Yet, only 12.5% of black and 32% of white Americans receiving home healthcare have an ACP. Despite the fact that the majority of older adults state that they would want to have discussions about care preferences, providers often remain reluctant to raise the issue when faced with prognostic uncertainty for patients who lack a clear terminal diagnosis or precipitating event. Data science methods, aimed towards discovering new knowledge in large datasets, shows tremendous potential for improving both individual and population-level quality outcomes. Nurse- scientists have made progress using these methods to improve care, increase satisfaction, and lower healthcare costs. The purpose of this study of community dwelling frail elders is to build a ?rules-based? prognostication model for clinical decision support for routine screening for mortality risk within 12 months to trigger home health nurses to identify those most at risk for dying. This data-science inquiry will use the Home Health Outcome and Assessment Information Set (OASIS), required by the Centers for Medicare and Medicaid Services to describe a variety of home healthcare measures. This study is unique because it develops and tests a novel methodological approach and because it is the only known study that will prognosticate death using a clinical outcomes measurement tool in current widespread use in this setting. The objectives of this study are aligned with the mission of the National Institute of Nursing Research (NINR) because it uses data- science methods to develop an evidence-based approach to improve communication, promote a shared understanding of prognosis, and facilitate patient and family-centered preferences for care at the end of life. The applicant will receive mentorship and training in data science methods from leading national and international experts in the field of health informatics, computer science, and mortality risk prognostication. The findings of this study will be foundational for an intervention study in which the predictive model is incorporated into clinical practice in an interoperable format to explore the impact that mortality risk prognostication has on the rates of advance care plans. Moreover, this study will be used to inform future research and to produce recommendations for clinical decision support to enhance bi-directional communication and collaboration between nurses, patients, families, and providers.